Exemple de CSV
| Term | Mentions |
|---|---|
| customer service | 142 |
| shipping delay | 98 |
| refund | 76 |
| packaging | 54 |
| out of stock | 41 |
| size chart | 33 |
| live chat | 29 |
| return label | 22 |
How to Create a Word Cloud
A step by step walkthrough for turning a list of words and counts into a published word cloud in PlotSet, from counting the terms to exporting the finished image.
Environ 8 minutesThis guide walks through how to create a word cloud in PlotSet from your own list of terms, not a demo dataset: sized on purpose, rotated on purpose, then embedded on a page or exported as an image once you're happy with it. What actually trips people up is not PlotSet itself, it is what happens before you ever open the template, so this guide spends real time on that step.
Five minutes if you already have a clean two-column list, word and count. Starting from raw text, survey answers, support tickets, a stack of comments, anything not yet counted, adds another five to ten. Treat the eight minutes below as the ceiling and counting as the part that decides whether the cloud is any good.
Les données dont vous avez besoin
Two columns, one row per term: the word or phrase first, its weight second. No running text, no duplicate terms, one header row.
The template wants terms already counted. If your source is raw text (a column of open-ended survey answers, say), run a pivot table or COUNTIF on it first: PlotSet sizes words by the number you give it, it does not read paragraphs and tally them for you.
Keep multi-word phrases as single cells, like "customer service" or "out of stock". If you split them into separate words before counting, you lose the phrase and end up with two vaguer terms instead of one useful one.
Remove stop words ("the", "and", "a") and any filler specific to your dataset before you count, not after. In support tickets, words like "issue" or "problem" are usually as empty as "the" and worth stripping too.
One row per term. If "refund" appears twice with two different counts, PlotSet will chart both rows rather than merging them, so combine duplicates in your spreadsheet first.
Les étapes
Le parcours exact dans PlotSet, d'un modèle vierge à un graphique terminé.
- 01
Count your terms before you open PlotSet
This is the step that actually determines the quality of the cloud. Take your raw text, whether it is survey responses, ticket subjects or comments, and produce a two-column count: term, then how many times it appears. A pivot table or a COUNTIF formula does this in a few minutes for most spreadsheets.
- 02
Strip stop words and dataset-specific filler
Remove "the", "and", "a" and their relatives, then look at your top ten terms again. Every corpus has its own version of filler: in product reviews it might be "item" or "product", words that appear everywhere and mean nothing on their own.
- 03
Open the Word Cloud template
From the PlotSet dashboard, start a new project and choose the Word Cloud template. It loads with sample terms and weights already in place, so you can see the exact two-column shape it expects before your own file goes anywhere near it.
Chaque guide commence ici. Choisissez la catégorie, puis le modèle : c'est la même grille quel que soit le type de graphique visé. - 04
Upload or paste your two columns
Import a CSV or .xlsx file, paste the rows straight from your spreadsheet, or connect a Google Sheet if you want the cloud to stay live as you keep counting. PlotSet reads the first row as headers automatically.
La vue données, ici sur un projet de graphique en barres. Importez un CSV, extrayez les chiffres d'une image avec 'Magic Import' ou synchronisez une Google Sheet. Le panneau est identique pour tous les types de graphique. - 05
Map the word column and the weight column
Bind the text column to the term and the number column to the size. This is usually correct on its own when your headers say something like "Term" and "Mentions", but check it if your sheet carries extra columns PlotSet had to guess between.
Les associations déterminent quelle colonne devient l'étiquette, laquelle fournit les valeurs et laquelle pilote la couleur. 'Auto Binding' les devine à partir de vos en-têtes : vérifiez-les plutôt que de les tenir pour acquises. - 06
Set the font size range
The minimum and maximum size decide how dramatic the contrast is between your smallest and largest term. Widen the gap if one or two terms genuinely dominate; narrow it if your counts are closer together and a wide range would just look noisy.
- 07
Set the rotation and colour
Rotation angles trade a tighter layout for slower reading, so 0 and 90 degrees only, or no rotation at all, is the easier choice to read. Pick a colour gradient; unless you are mapping a second value to it, treat it as decoration and keep it simple.
L'onglet 'Preview' pour un graphique Word Cloud. Tous les réglages de style se trouvent dans le panneau entouré en rouge : apparence, en-tête, pied de page et les options propres à ce type de graphique, chacun dans un onglet distinct. - 08
Publish, embed or export the cloud
Publish for a live, shareable cloud, grab the embed code for a page, or export a PNG, SVG or PDF for a deck. Choose SVG if the image is going into print or will be scaled up, since the words stay sharp instead of pixelating.
Partez du modèle Word Cloud avec des données d'exemple déjà chargées, puis remplacez-les par les vôtres.
Ouvrir le modèlePublier, intégrer ou exporter
La dernière ligne droite est la même pour tous les types de graphique : 'Share' transforme le graphique en lien public actif et en code d'intégration, 'Export' l'enregistre sous forme de fichier statique.
Share : lien public et code d'intégration
Activer le code d'intégration et le lien public revient à publier le graphique. Le lien peut être envoyé à qui vous voulez sans risque, et le code d'intégration s'insère dans n'importe quelle page sous la forme d'une iframe qui conserve les infobulles au survol et l'interactivité. Si vous modifiez le graphique par la suite, cliquez sur 'Recreate' pour répercuter ces changements sur le lien comme sur le code.
Export : SVG, PNG ou JPG
Le bouton 'Export' génère une copie statique à la taille dont vous avez besoin. Le SVG reste net à toutes les échelles et peut encore être modifié dans un outil de design ; PNG et JPG sont les choix sûrs pour les présentations et les documents. C'est le multiplicateur d'échelle qui vous donne une image nette sur un écran retina.
Conseils de style pour un Word Cloud
Keep phrases as single terms
"Climate change" counted as one term carries a different, more useful signal than "climate" and "change" counted apart. Decide which phrases matter to you before you count, not after you have already split everything into single words.
Turn rotation down before you turn it up
Free-angle rotation packs more words into the same space, but every rotated word costs the reader a beat to read. Start with rotation off, look at how much white space is left, and only add angles if the layout genuinely needs them.
Cap the list around fifty to a hundred terms
Somewhere past a hundred, the smallest words are too tiny to read and are only there to fill gaps. Trim your count list to the terms that carry the story, usually the top fifty to a hundred by weight, before uploading.
Publish the counts somewhere, even if not on the cloud itself
A cloud cannot tell a reader that "refund" appeared 76 times versus 74, and it should not try to. Keep a caption, a footnote or a linked table with the real numbers for anyone who wants to check the sizing rather than just feel it.
Erreurs courantes en créant un Word Cloud
The biggest word on the chart is "the" or "and"
A stop word made it into your count list before you filtered it out. Go back to the source spreadsheet, remove the stop words and any dataset-specific filler, and re-upload rather than trying to fix it inside PlotSet.
The same word shows up twice at two different sizes
This means your count list has duplicate rows for one term, usually from inconsistent capitalization ("Refund" versus "refund") or a trailing space. Merge the rows in your spreadsheet, ideally with a TRIM and a case-normalizing formula, then re-upload.
Every word comes out roughly the same size
Your counts are too close together for font size to show any real difference, which usually means your top terms are all within a factor of two of each other. Either widen the size range in the chart settings, or accept that this dataset is better shown as a ranked bar chart.
En savoir plus sur le Word Cloud
Vous n'êtes pas sûr que le Word Cloud soit le bon choix ? Découvrez ce que c'est, quand l'utiliser et quand l'éviter.
Word Cloud : questions fréquentes
Count your terms into two columns first, word and weight, since PlotSet sizes words by the number you give it rather than counting raw text itself. Then open the Word Cloud template, upload or paste the file, confirm the column mapping, and publish.
Créez votre Word Cloud maintenant
Gratuit pour commencer avec plus de 300 modèles. Aucune carte bancaire requise.